Home > Blog > Fabricating Nature

Nature is unbelievably complex. Animals and humans have evolved in this world without in-built digital systems. Our minds have evolved in ways that allow us to make sense of our environment, so that we can abstract and categorize the things that we encounter and experience. As such, we have an in-built tendency to represent the world in terms of simple, clearly identifiable boundaries of space and objects. When we create objects ourselves, we continue to follow this path without even realizing it. Traditional manufacturing and design assumes that each object, or each independent part of a larger object, is made of a single homogeneous material. This makes human-made objects clearly stand apart from nature. A tree, for example, is not made from a single material and nor does it have 'parts' that are as easily discernible as we would think - if we look up close we can see that roots blend into the trunk, which blends into branches which blend into twigs, which blend into leaves. Our abstractions are useful for recognition and categorization, but they do have limitations when it comes to creating or recreating complex objects like this.

Historically then, we've interacted with nature through powerful but reductive simplifications and approximations - the way we look at it, the way we model it and the way we attempt to reproduce it. Cheap computing power is now extending our capabilities, putting us in a far better position to understand the complexities of nature. We have better control over matter and can design and fabricate a whole new class of human-made objects. These objects offer more localized, dynamic, sustainable and natural interactions with the world. Unfortunately, we've hit a stumbling block. The current generation of digital design and fabrication systems have failed to fully capitalize on the raw computational power that is available to us. We still can't create objects that offer a comparable wealth of detail, complexity and combination of materials that we find in nature. Here is an example of how a typical user might model a watermelon using 3D modeling tools and a real watermelon. There is little to compare other than the rough outside shape. How can we truly represent a slice of watermelon digitally?

Let us take a very basic example, one that is even human made, a glass marble. Marbles are simple children's toys, but how can we exactly model their construction?

The only real 'surface' here is the outer shape, but modelling it with polygons will always leave it faceted. Perhaps we can try to overcome this inaccuracy and using a parametric surface, but can we efficiently represent the minute chips made by numerous games of marbles? Still, a spherical surface won't allow us to define what's inside. Perhaps each of those different colors could be a separate part? If we look closely at the swirls we can see that they don't have sharply defined edges. The colors mix and blend together throughout the interior. Our traditional approach of modelling with surfaces falls apart completely. Our entire approach is wrong because this is a problem that requires real volumes with no neatly defined 'parts'.

Is this a fair example? Would anyone actually want to recreate a watermelon or 3D print a glass marble? Perhaps not, but what about a human organ such as a kidney. That's also a challenging volumetric problem and one that would be of huge benefit if we could design and fabricate them as required.

So, the way we think about the world allows us to make sense of it, but doesn't necessarily let us reproduce it. Existing digital systems have followed our natural way of thinking and have led us to an impasse. They are imprecise and fundamentally incapable of accurately representing real objects. If we want to move forwards, we will need to take a very different approach.

Share this post


Comments (1)

The problem you describe to fabricate nature is not geometry alone, it is about lighting. You want to fabricate the structure of the watermelon, but without the actual transmissive properties of the material how will it look like a watermelon under natural light.

To model that nice translucent sheen, computer graphics does not depend on the geometry, but a complex shading model for describing light bouncing around. In contrast, if the geometry was modeled at a that extreme level of infitismal surfaces, it would not produce the desired visual effect unless you used the correct material information. A shading approach is more computationally feasible than extreme geometrical modeling (yes, computer graphics cheats!).

I would also argue that polygons are justified if they match the resolution needed. Most rendering processes undergo digitisation at some stage, often ray tracer will break the surface into enough polygons so that there is no perceptual loss of visual detail. Moreover, numbers in a computer are an approximation in binary form. It is convenient that enough bits, such as a 64 bit floating point number, will produce enough decimal places so that no one will notice that approximation. Its just a matter of getting to the resolution level that you need.
John , December 22, 2012

Write comment

smaller | bigger
security image
Write the displayed characters